Scale qual with quant: Comparing AI-led and human-led conversations